Target's Marketing Strategy: Pregnancy Prediction Algorithm

Abstract

This report aims to highlight how the retail stores are using big data and predictive analytics to know their customer’s need and do personalized advertising. The report focuses on the enhanced marketing operation of US retail giant Target through the use of big data. Target was one of the first retailers to use data-driven prediction to do personalized marketing. They developed an algorithm to predict customer behavior and it was so accurate that Target was given the title “the store who knows you are pregnant before you do”. When the company’s successful strategy of using big data backfires they should work around the strategy to make it acceptable with a certain twist. Target’s pregnancy prediction algorithm backfired when angry dad of teenage daughter verbally bashed the store asking if they’re encouraging his little daughter to get pregnant. Target modified their prediction algorithm, when they knew even when there haven’t been any privacy breaches, there could be ethical implications and it can hurt customer sensitivities, by adding certain subtlety.

Introduction

Retail stores have challenges that are day to day or long-term which should be managed at the same time. They generate vast amounts of data across various operative divisions. For retailers, data can be an asset only if they are able to make sense of it. Target Corp. is the second-largest retailer in America, providing a convenient one-stop shopping destination. It has more than 1800 stores all over the United States with more than 80 million customers visiting stores each year. It accounted around 71 billion revenue in 2018 and is just behind Walmart. Target is constantly examining how to leverage innovative technology, analytics, and data to optimize the business model. From the marketing point of view, consumers are willing to change their habit and buying decision when they’re moving to a new phase of life i. e. when some life events are happening e. g. getting married, moving to a new city etc. So if a company is able to determine the life event and send the personalized advertisement, they have a higher chance of earning their loyalty as long-term customers. To capitalize on this theory Target hired Andrew Pole who had to look into daily customer buying behavior and predict the pregnancy of women so they can push personalized pregnancy-related product suggestion to those women. Soon after, Target began sending discount coupons based on the rating of pregnancy prediction algorithm. This strategy backfired when angry father of 17-year-old teen walked into Minneapolis store and questioned if they’re encouraging his daughter to get pregnant by sending her coupons for baby clothes and cribs. The manager who didn’t know why this algorithm failed, obviously, apologized and also called back a few days later to apologize. However, the father apologizes saying there have been some activities in his house of which he was completely unaware and her daughter is in August. All the long ‘pregnancy prediction algorithm’ was working perfectly and even was ahead of a family member. But this raised concern regarding the ethics and human sensitivities for which they had to modify algorithm adding certain subtlety. Data gives a power to companies if utilized effectively but there’re certain things that are within the boundary of law but still needs to be addressed correctly in order to sustain.

Target’s Predictive Analytics

The trend of collecting customer’s information is not new for any companies. In the case of retail stores, they collect the large amount of information on the customers walking to their store or visiting their website, social profile etc. Retailers can capture the value of data using predictive analytics by analyzing customers based on metrics such as demographics, age, gender, income etc. and guide the customer to a new product. Machine learning models are trained in historical data that can generate accurate recommendations, Analyzing purchase history, and the company can predict what the customer is likely to purchase next. For “pregnancy prediction algorithm” Target also collected a vast amount of data on consumer and assigned them unique Identifier (Guest ID number) whenever possible. That ID was used to keep a tab on everything they buy and also other information like demographics, marriage status, time to reach store etc. Whenever customer used a credit card, discount coupon, email for a refund, open advertisement email etc. they record it and assigned to Guest ID. Pole looked at historical buying data for all the ladies who had signed up for Target baby registries in the past in conjunction with data bought from an external source. Women were buying larger quantities of unscented lotion, supplements like calcium, magnesium etc. around the 4th month. The sudden increase in buying of scent-free soap, extra-big bags of cotton balls, hand sanitizers and washcloths indicates near delivery date, Crawling on these data Pole’s team identified 25 products that were purchased in combination during pregnancy which he used to assign “pregnancy score”. Depending upon the category of product purchased, they could predict pregnancy’s trimester, and thus approximate due date. So Target started sending coupons for baby items to customers according to their pregnancy scores. After the incident with father of a teenage daughter, Target identified that even if the company is following the law, it could freak out its customer if the strategy is not analyzed from the various perspective of the customer.

Workaround Strategy

It is not just enough to collect vast amount of data and create business strategy based on it. Beside identifying the right customer data and developing precise prediction algorithm, companies must take into consideration the consumer’s sensitiveness and make a consumer feel that their privacy is intact and haven’t been breached. When news of teenage pregnant went public target realized that using data to predict a woman’s pregnancy can harm the public relation. As a turnaround strategy Target used negative score approach. For this approach they started mixing all the things that for sure pregnant women never buy and included in between the advertisement for pregnant women. This made the advertisement look random and give an impression that the product was chosen by chance. For example putting a refrigerator advertisement next to diaper made and ad look random. As long as the consumer feels their privacy has not been breached they’d behave normally and there is greater ease in converting them to loyal consumer. For this companies have to add subtlety in the approach they deal with customer by using their data which they shared unknowingly or knowingly.

Technologies Solution

To manage a huge amount of data there are various technologies available which are widely used by various companies. Solutions to big data problems are provided by tools such as Hadoop, Map reduce, Spark etc. There are also software solutions providing companies who provide a stack of big data tools as per the need of business and few companies provide support in managing data. Target has its own human resource to work on the data related solution, with each division equipped with data analysts and data scientists. It has also partnered with various companies to get efficient data analytics solution. The company has partnered with Teradata to warehouse its data assets. Similarly, it has a partnership with Micro Strategy which provides data visualization and performance monitoring services. It has also a partnership with Axiom, a company that collects, analyzes and sells customer and business information used for the targeted advertising campaign and Epsilon who provides data-driven marketing services.

Conclusion

The retail industry continues to accelerate rapidly, so they have to find out the best use case for the available data. Though commercial benefits of big data is huge, it also has some potential downside that needs to be addressed properly. Volume of data is not everything but the value that could be generated form it speaks volume. The value generated form data should be beneficial for both providers and consumer and dissatisfaction of either end breaks the links which destroys the possible value. Target used its consumer’s information to help them choose the item they’d possible need so as to increase the loyalty. They played with data and developed what was supposed to be accurate algorithm. But they failed to recognize and incorporate the human sensitiveness factor in their algorithm and hence were backfired. If a company is trying to predict a very personal and sensitive situation, they should try to be subtle in the way the use the information and communicate with the customer. Before implementing any predictive analytic project, company should also think about the ethical implication and how the public would react to if they’re to find out how this project is working. Data gives the power to do many thing but that doesn’t mean the company can do anything they can with this power.